[CLOSED] - Stumptown Miners - Avalon PCB Assembly - West Coast USA

[tolan77]

Tolan, I'll send an email to Colin asking about the six chips. I'm sorry you haven't received them yet.

Any word? I have yet to receive anything in the mail nor a reply via pm since Colin's latest update via email that you posted here. Curious what's happened since.

[ryepdx]

Tolan, I'll send an email to Colin asking about the six chips. I'm sorry you haven't received them yet.

Any word? I have yet to receive anything in the mail nor a reply via pm since Colin's latest update via email that you posted here. Curious what's happened since.

Hey Tolan,

Sorry about that. I'd told Colin to hold on sending those chips and to reply with a status update since you'd mentioned your decision hinged on whether the K16 design was ready to go or not, IIRC. I've just emailed him telling him to go on and send the chips.

Also, here's a status update from Colin a few days ago:

...
Now for the parts, I'm thinking we may do better just building more K16s for other people with chips or re-purposing the parts on a Gen2 design.  Not sure.  Just that kitting them out in 'refunds' isn't really doable.  If I do get a Gen2 design going I may buy them myself, but that's really speculative at this point.

As for the K16, I'm not sure at this point.  They are hashing, but with ~25% HW that just doesn't seem to be getting fixed.  It looks like there is a problem with the PIC and USB and CGMINER, but I really can't tell if its fixable on the current PCB.  It is taking way longer then I figured it would which is a bad sign... The PIC seems to be underpowered for the task at hand.
...

And another from him 23 hours ago:

...
I'm really sure the problem with the current K16 is firmware.  The group seems to be having a hard time as well with CGminer itself.  It all revolves around the USB being the trouble spot.  My hunch is the little PIC can't handle feeding the ASIC's and dealing with USB traffic at the same time.  Should be able to fix it, but who knows how experience the guys working on it are.  It isn't trivial code that's for sure.  USB is a dragon.

I've already written a simple stand alone stratum client.  I'll try to write K16 firmware from scratch that doesn't use USB or CGminer.  This may at least help the others find the issue.  Right now it seems a struggle to debug with CGminer/USB being so complex.  I want a really simple test setup.  This will lead right into doing my own Gen2 board as well.

I'll be back home later this week and should be able to get my test client running with the K16.  If that works, and I suspect it will, we'll be able to figure out some kind of solution.
...

[ryepdx]

Woke up this morning to these two emails from Colin:

They found the firmware bug!!  So far two people report 0 (Z.E.R.O.) HW errors and over 5gh/s.

It's really late, so I'll confirm mine in the morning.

I've got a few work items to deal with tomorrow am, but after that (and I confirm myself) I'll start running boards.  Also need to order the replacement parts.

It was a dumb bug.  Just a matter of time before it was found.

Ok.. Couldn't wait   I can confirm no errors on my unit.  We're golden.  It's hashing right now.

So those of you still in, it looks like we're good to go!

[hephaist0s]

Fantastic! That is great news!

[DPoS]

Are they testing these boards with Gen2 chips yet?

[davecoin]

Afaik there are no gen 2 chips in hand yet, other than those shipped to board designers for the contest. 

[hephaist0s]

From tradehill.com:

We are pleased to announce the launch of our next auction. BitSyncom is auctioning their Avalon ASIC A3255-Q48 chips: the second generation SHA256 Processor. BitSyncom will be auctioning off 20 reels of 500 chips per reel. Each chip outputs 1.5-1.6 GH/s at a standard 0.9V, with 2.05W consumption. The minimum bid for a single reel will be 24 BTC.

For more detailed information on the A3255-Q48 mining chip, please see Avalon’s Datasheet. This product has already cleared customs and is ready to ship upon the close of the auction.

Bidders who have deposited bitcoin in their Tradehill Auction accounts may begin bidding as of Friday November 1, 2013, 1:00 PM EST. The auction will close on Tuesday November 5, 2013, 2:00 PM EST.

All bidders that have submitted identity documents since the close of the prior auction should have received a response from Tradehill Auction Support either indicating that their account has been approved or detailing additional documentation needs.
For support related to your auction account, please contact auctionsupport@tradehill.com.

[DPoS]

That's some crappy stats

[hephaist0s]

That's some crappy stats

If they work with the existing board design for the gen 1 chips, then they're great stats.

[DPoS]

That's some crappy stats

If they work with the existing board design for the gen 1 chips, then they're great stats.

that would be nice, two K16's would be around 50 Gh

but if the lead time to tweak things takes 3 more months it just becomes a scholastic endeavor

[ryepdx]

Quick update:

The firmware on the miners does not yet allow for chaining K16s together using "klego" adapters, nor does it allow for updates via USB. If you take receipt of your miners now and you want to be able to chain your miners together sometime in the future, you will eventually need to flash them with a PIC ICD3 or PicKit programmer. If you aren't comfortable with doing that, I will gladly host your miners for the cost of electricity until the firmware is updated and I can flash them myself.

Colin has begun cranking out miners, so you will be receiving yours shortly unless you decide you want me to hold on to them for the time being. Please PM me if that's the case.

Thanks,

Ryan

[Bicknellski]

+1 Ryan.... nice work.

Glad to see the finish line on these I bet?

[ryepdx]

Oof. Not so fast. I just got this email from Colin:

Unfortunately I've got some troubling news.  After assembling a number of miners I'm getting odd test results.  All of the miners hash to some degree, but at very differing rates with most reporting hardware errors now and then.  I've been working for days now trying to figure out whats going on.  My guess is some of the ASIC's are not contributing or are not hashing properly.

I'm not really sure what to do next.  Right now, I have 5 fully built with another 10 assembled but not re-flowed.  I stopped assembly and reflow when I started to get odd results testing.

Of the 5, 1 works perfectly (my first unit), another almost as good, and the others less then ideal.

Needless to say I don't want to keep going until I can figure out the problem, but I'm not sure its worth it at this point ?  I'm burning a ton of time on this project, and while interesting there was never a budget for it.

To make matters worse, I've got to head out of town Wednesday morning, so I'm running out of time.  I'll be back in two weeks, but trying to ship these in a timely manner is looking unlikely.  It's bad timing but I never expected to still be working on these.

At this point, there's a couple of options:

1. Pull the plug.  I can send the few working units and all the parts.  I'd be willing to refund 75% of the assembly cost.

2. Keep at it.  I can't do much before leaving, but can attempt to debug the units when I return.  I'll send the two units that seem to work the best asap.

Right now, most of the parts/asics have not been reflowed, meaning I can take them off the assembled PCB's, clean them and try again or just send them back.

To figure out whats not working I'll need to setup a test environment and probe the boards until I find a dud ASIC and then try to make it work.  ktest may help with that.  Unfortunately the firmware as is contains no diagnostic code or report of ASICs not responding.

The biggest issue is time.  Time spent and time to get them delivered.

Not the news I wanted to give you.  Let me know your thoughts asap and we can figure out what to do next.

Since these are your miners, I thought I should get some feedback on how you want me to proceed now. I don't know how many hardware errors there are, nor do I know how much of a performance hit the miners are taking. I emailed Colin asking for some hard numbers to help with the decision-making.

Thoughts on this?

[Bogart]

Tough call.  Precise performance numbers may help in deciding.

I'm inclined to take delivery even if they're not fully up to snuff.  Some hashpower in the hand is better than none in a pile of parts IMO.

[ryepdx]

Colin with hard figures:

The unit that works perfectly runs at ~71wu, the others are 50, 35, 30 and 20.  Those are relative and approximate since they change a little as it runs.  71wu/s is the full 5.5gh/s.

[Bogart]

I guess I'll take delivery of mine, even if they're underperforming.

At least I can hash with them while searching/waiting for a solution.

[tolan77]

Please just send mine out unless you're planning to make any headway on it soon. It's really hurting that I put about 8btc into these chips and this project :p

[tolan77]

Hello? Any word?

[ryepdx]

So far I've received and shipped on Bogart's miners. Nothing yet on yours, tolan77. Sorry. I emailed Colin again today and will hopefully hear back soon.

[tolan77]

So far I've received and shipped on Bogart's miners. Nothing yet on yours, tolan77. Sorry. I emailed Colin again today and will hopefully hear back soon.

:/ I ordered 4xK16 miners from Zephirs Batch 1 of chips is there any chance I'm actually going to be receiving any of these? If you don't mind me asking how many orders with chips batch 1 Zefir did you have besides Bogart?